Detection of Plasmodium berghei infected Anopheles stephensi using near-infrared spectroscopy

Abstract

Abstract Background The proportion of mosquitoes infected with malaria is an important entomological metric used to assess the intensity of transmission and the impact of vector control interventions. Currently, the prevalence of mosquitoes with salivary gland sporozoites is estimated by dissecting mosquitoes under a microscope or using molecular methods. These techniques are laborious, subjective, and require either expensive equipment or training. This study evaluates the potential of near-infrared spectroscopy (NIRS) to identify laboratory reared mosquitoes infected with rodent malaria. Methods Anopheles stephensi mosquitoes were reared in the laboratory and fed on Plasmodium berghei infected blood. After 12 and 21 days post-feeding mosquitoes were killed, scanned and analysed using NIRS and immediately dissected by microscopy to determine the number of oocysts on the midgut wall or sporozoites in the salivary glands. A predictive classification model was used to determine parasite prevalence and intensity status from spectra. Results The predictive model correctly classifies infectious and uninfectious mosquitoes with an overall accuracy of 72%. The false negative and false positive rates were 30 and 26%, respectively. While NIRS was able to differentiate between uninfectious and highly infectious mosquitoes, differentiating between mid-range infectious groups was less accurate. Multiple scans of the same specimen, with repositioning the mosquito between scans, is shown to improve accuracy. On a smaller dataset NIRS was unable to predict whether mosquitoes harboured oocysts. Conclusions To our knowledge, we provide the first evidence that NIRS can differentiate between infectious and uninfectious mosquitoes. Currently, distinguishing between different intensities of infection is challenging. The classification model provides a flexible framework and allows for different error rates to be optimised, enabling the sensitivity and specificity of the technique to be varied according to requirements.

Keywords

• Vector borne diseases
• Vector control monitoring
• Anopheles stephensi
• Plasmodium berghei
• Near-infrared spectroscopy
• Machine learning